Codex Pro: 100% weekly credit exhausted by "Codex exec" — what is it and how do I stop it?

Hi community,

I’m a Codex Pro subscriber, and I’ve run into a very puzzling situation I hope someone can help with.

The problem: My entire weekly credit allowance has been completely exhausted, but almost entirely from something called “Codex exec”, which isn’t something I recognize, and didn’t knowingly enable, and can’t find any way to disable.

What’s confusing me:

  • My actual development workflow this week hasn’t changed at all — I’ve just been using the standard GPT-5.4 xhigh, interface as usual. I’ve never ran into either the 5hour limit, or the weekly limits, despite relying on flagship xhigh models since the beginning.

  • Looking at the weekly credit usage breakdown (screenshot attached), 100% of the consumption is attributed to “Codex exec” — none of it comes from the 5-hour compute limit I expected:

  • This started on Tuesday, the 17th of March, and has continued for multiple days

What I’ve already tried:

  • Logged out of the Codex desktop app (which I had been using separately on some repos)
  • Logged out of Codex CI within Cursor IDE
  • Ensured that Github workflows like PR reviews aren’t using credits
  • Despite all of these, I’m still seeing ongoing credit consumption from “Codex exec” across multiple days

My questions:

  1. What exactly is “Codex exec”? I can’t find clear documentation on this billing category
  2. How did it get enabled without my knowing?
  3. How do I turn it off / stop the credit drain?
  4. Has anyone else experienced a similar unexpected spike from this source?

I’ve searched the forum and can’t find similar reports. Any help would be greatly appreciated — this has wiped out my weekly credits with seemingly no corresponding work done on my end.

This also just happened to me. Similar to you, I am using a workflow that is unchanged. exec suddenly depleted my budget. Did you get any additional insight?

I have edited to attach the below image showing continuous work on two projects with a sudden increase of this unknown process yesterday. Given there is no documentation or information available, it would be nice to hear something from OpenAI.

Same here! Can we get some resposne plese? This is ultimately making the codexunpredictable and unusable!